You are on page 1of 11

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/234862398

Carbon Metal Fluoride Nanocomposites High-Capacity Reversible Metal Fluoride


Conversion Materials as Rechargeable Positive Electrodes for Li Batteries

Article in Journal of The Electrochemical Society · October 2003


DOI: 10.1149/1.1602454

CITATIONS READS
412 991

4 authors, including:

Frederic Cosandey Nathalie Pereira


Rutgers, The State University of New Jersey Rutgers, The State University of New Jersey
157 PUBLICATIONS 3,784 CITATIONS 113 PUBLICATIONS 5,125 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Nathalie Pereira on 11 November 2014.

The user has requested enhancement of the downloaded file.


A1318 Journal of The Electrochemical Society, 150 共10兲 A1318-A1327 共2003兲
0013-4651/2003/150共10兲/A1318/10/$7.00 © The Electrochemical Society, Inc.

Carbon Metal Fluoride Nanocomposites


High-Capacity Reversible Metal Fluoride Conversion Materials
as Rechargeable Positive Electrodes for Li Batteries
F. Badway, F. Cosandey, N. Pereira, and G. G. Amatucci*,z
Department of Ceramic and Materials Engineering, Rutgers-The State University of New Jersey, Piscataway,
New Jersey 08854, USA

The structure and electrochemistry of FeF3 :C-based carbon metal fluoride nanocomposites 共CMFNCs兲 was investigated in detail
from 4.5 to 1.5 V, revealing a reversible metal fluoride conversion process. These are the first reported examples of a high-capacity
reversible conversion process for positive electrodes. A reversible specific capacity of approximately 600 mAh/g of CMFNCs was
realized at 70°C. Approximately one-third of the capacity evolved in a reaction between 3.5 and 2.8 V related to the cathodic
reduction reaction of Fe3⫹ to Fe2⫹. The remainder of the specific capacity occurred in a two-phase conversion reaction at 2 V
resulting in the formation of a finer Fe:LiF nanocomposite. Upon oxidation, selective area electron diffraction characterization
revealed the reformation of a metal fluoride. Evidence presented suggested that the metal fluoride is related to FeF2 in structure.
A pseudocapacitive reaction is proposed as a possible mechanism for the subsequent Fe2⫹ → Fe3⫹ oxidation reaction. Preliminary
results of FeF2 , NiF2 , and CoF2 CMFNCs were used in the discussion of the electrochemical properties of the reconverted metal
fluoride.
© 2003 The Electrochemical Society. 关DOI: 10.1149/1.1602454兴 All rights reserved.

Manuscript submitted November 1, 2002; revised manuscript received April 15, 2003. Available electronically August 15, 2003.

The Li-ion battery is the premiere high-energy rechargeable en- and nitride materials. The output voltage of primary conversion re-
ergy storage technology of the present day. Unfortunately, its high actions are rooted in basic thermodynamics and are well understood.
performance still falls short of energy density goals in applications In general, an increase in the Me-X bond ionicity results in an in-
ranging from telecommunications to biomedical. Although a number crease in the output voltage associated with Reaction 1. Therefore
of factors within the battery cell contribute to this performance pa- the output voltage of Reaction 1 would expect to increase through
rameter, the most crucial ones relate to how much energy can be the highly covalent metal nitrides and sulfides, to the metal oxides,
stored in the positive and negative electrode materials of the device. through the inductive effect polyanions 共metal phosphates, metal
The positive electrode of Li-ion batteries is dominated by the borates兲, and finally to the highly ionic metal fluoride and metal
layered Li intercalation compound, LiCoO2 . 1 LiCoO2 has a practi- chloride halogens.
cal reversible specific capacity of 150 mAh/g. Alternative electrode Metal fluorides have been largely ignored as reversible positive
materials such as compounds and solid solutions consisting of electrodes for rechargeable lithium batteries. This is due to their
LiNiO2 2 and LiMn2 O4 3,4 have been introduced in the past. Although insulative nature brought about by their characteristic large bandgap.
the capacity of these materials does not exceed that of LiCoO2 to a Iron trifluoride (FeF3 ) was first reported as showing limited electro-
great extent 共⬍250 mAh/g兲, they are lower in cost and the latter is chemical activity by Arai et al.11 Arai reported a capacity of 80
more environmentally acceptable. For the past decade there has been mAh/g from FeF3 in a discharge voltage region from about 4.5 to 2
an extensive effort to search for new positive electrode materials. V involving the Fe3⫹ to Fe2⫹ redox transition. The poor electronic
Current focus is on layered manganese compounds of the general conductivity combined with a questionable ionic conductivity re-
formula LiMnO2 5 and phosphate materials of the general formula sulted in the disparity between the observed reversible specific ca-
LiMePO4 6 and Li3 Me2 (PO4 ) 3 7 where Me is a transition metal. Al- pacity of 80 mAh/g and the theoretical (1e⫺ transfer兲 capacity of
though operating at a lower voltage and close to the same capacity 237 mAh/g. Since then, the fluorides have been discarded as useful
as present day LiCoO2 , these materials are of interest due to their reversible positive electrode materials.
low cost and safety. However, little new ground has been revealed in Our group has recently introduced the use of carbon metal fluo-
the quest for positive electrode materials of higher energy density. ride nanocomposites 共CMFNCs兲 to enable the electrochemical ac-
The fundamental route to attaining the highest specific capacity tivity of metal fluorides.12 The concept was rooted in the fact that
of an electrode is to utilize all the possible oxidation states of a nanosized crystals have a high total material volume on the surface.
compound during the redox cycle. This can be done by way of a This surface contains a number of defects which can contribute sub-
reversible conversion reaction which proceeds stantially to enhanced electronic and ionic activity 共for example, see
Maier13兲. In addition to increased activity of the metal fluoride
nLi⫹ ⫹ ne⫺ ⫹ Men⫹X ↔ nLiX ⫹ Me 关1兴 brought about by the drastic reduction in crystallite size, we have
electrically connected the grains of each through the use of highly
Previously, such reversible conversion reactions were only shown to conducting carbon. Our initial studies have shown a major improve-
be active for low-potential materials suitable for use as negative ment in the electrochemical activity of FeF3 in the 4.5-2.5 V region
electrodes. This was first demonstrated by Tarascon’s group for the such that we could recover 99% of the FeF3 theoretical capacity
oxide and sulfide chalcogenides8 and later by our group for the 共235 vs. 237 mAh/g兲 in the 4.5-2.5 V region with a total CMFNC
transition metal nitrides.9 However, limited reversible conversion of specific capacity of approximately 200 mAh/g. The capacity and
the Cu with ␣-CuVO3 has been noted by Kamiya et al.10 voltage was consistent with the reduction
The quest to increase the specific capacity of the positive elec-
trode would require adoption of reversible conversion to positive Li⫹ ⫹ e⫺ ⫹ Fe3⫹F3 → LiFe2⫹F3 关2兴
electrode materials, necessitating at least a 1 V increase in output
voltage with respect to the previously demonstrated chalcogenide, We also recently presented evidence that FeF3 CMFNCs offered
excellent reversible specific capacity through a second reaction oc-
curring at 2 V. The combined specific capacities resulted in an ex-
* Electrochemical Society Active Member. ceptional total capacity of approximately 600 mAh/g. We reported
z
E-mail: gamatucc@rci.rutgers.edu that this metal fluoride reaction was due to a reversible fluoride-
Journal of The Electrochemical Society, 150 共10兲 A1318-A1327 共2003兲 A1319

based conversion reaction and could be utilized for positive elec-


trode technology. Herein we present details of this reaction along
with preliminary data of CMFNCs based on the transition metal
difluorides: FeF2 , NiF2 , and CoF2 .

Experimental
High-energy milling.—CMFNCs were fabricated by the high en-
ergy milling of R3̄C FeF3 共Alfa兲, and either expanded graphite 共Su-
perior兲, carbon black 共MMM Super P兲, or activated carbon 共Norit
A-supra兲. Stoichiometric mixtures 共metal fluoride: C, 85:15 wt %兲
were placed inside a hardened steel milling cell along with hardened
steel media. All cell assembly was done in He atmosphere. The
milling was performed for the designated times in a high-energy
mill 共Spex 8000兲. As-formed CMFNC samples were removed from
Figure 1. XRD patterns of FeF3 :C CMFNCs fabricated by high-energy
the milling cell in a He-filled glove box. Unless otherwise noted,
milling with 15 wt % activated carbon for 30 min and 1 h compared with
85/15% FeF3 :C CMFNCs fabricated by high-energy milling for 1 h manual mixing. 共*兲 Marks impurity peaks related to FeF3 -3H2 O.
with activated carbon were utilized for general characterization.
Physical characterization.—XRD.—The structure of the materi-
als was characterized by X-ray diffraction 共XRD兲, utilizing a Scin- Pile 共Biologic兲 or Maccor battery cycling systems. Cells were cycled
tag X2 diffractometer with Cu K␣ radiation. In situ XRD was car- under a constant current of 7.58 mA/g at 22°C, unless noted other-
ried out in a custom in situ cell developed by our group and utilizing wise.
a Be window, details of which will be published at a later date. The
electrochemical cell was cycled at a constant current of 7.58 mA/g. Results
Three-hour scans were utilized resulting in a ⌬x of 0.115 共x in
Lix FeF3 ) per scan. Ex situ XRD was accomplished by disassembling FeF 3-C CMFNCs.—Physical characterization.—FeF3 -based
the electrochemical cell in a He-filled glove box and rinsing the CMFNCs were fabricated by high-energy milling FeF3 and activated
electrode in dimethyl carbonate 共DMC兲. The dried electrode was carbon, carbon black, or expanded graphite for 10, 30, 60, 240, and
then placed on a glass slide and covered with Kapton film using a 360 min in a 85-15 wt % ratio. The various carbons were chosen for
seal layer of vacuum grease. their contrasting properties. Activated carbon is a meso- and mi-
Transmission electron microscopy.—The material microstructure croporous carbon with a surface area of 1700 m2/g, carbon black is
was analyzed by transmission electron microscopy 共TEM兲 utilizing networked carbon with excellent electrical connectivity, and ex-
a Topcon OO2B and a Phillips EM30 microscope. Dark field 共DF兲 panded graphite is an acid-treated, partially exfoliated graphite with
imaging techniques were used to image the size and distribution of excellent electric conductivity. All samples were analyzed by XRD
the various phases. In addition, selected area electron diffraction to study phase development. XRD patterns of FeF3 -activated carbon
patterns 共SAED兲 from various areas were obtained to determine the CMFNCs fabricated for 30 min and 1 h, compared with a macro-
structure of the phases present. SAED patterns were analyzed using composite fabricated by manually mixing 15% activated carbon
Process Diffraction software.14 Ex situ high-resolution TEM 共HR- with 85% FeF3 , are shown in Fig. 1. Substantial and systematic
TEM兲 was performed on samples in various states of electrochemi- broadening of the Bragg peaks occurred with increasing milling
cal conversion to determine crystallinity and crystallite size. TEM times, as is clearly evident in the comparison of the 共012兲 Bragg
samples were prepared by dispersing the milled powder in tetrafluo- peaks. The XRD data reveals that FeF3 remained as the only signifi-
rocarbon and releasing a few drops of the liquid on a carbon film cant phase developed after milling up to 1 h; however, a small
supported on a copper grid. Ex situ TEM was accomplished by quantity of FeF3 -3H2 O is present in the raw FeF3 which dissipates
disassembly of coin cells containing powder. The powders were after milling. The small amount of hydrated FeF3 may be residual
rinsed in DMC, and grids were prepared as described previously from the dehydration process which can be utilized to form
except that dispersion was accomplished with the use of DMC. All commercial-grade FeF3 . The destruction of this small amount of
preparation was accomplished in an anhydrous atmosphere. The hydrated FeF3 phase by the high-energy milling is not surprising.
grids were sealed under helium and transferred to the TEM chamber. Localized heating encountered by the sample during the milling pro-
The final transfer to the TEM chamber consisted of less than 30 s cess is most likely beyond that of the dehydration temperature.
exposure to ambient atmosphere. Milling the activated carbon sample for 4 h 共Fig. 2a兲 results in
Electrochemical characterization.—Following the removal of the reduction of the Fe3⫹F3 phase through defluorination, as evi-
CMFNC samples from the milling cell in the He-filled glove box, denced by the major phase of Fe2 ( 3⫹2⫹) F5 and minor presence of
the samples were sealed in a vial and transferred to a dry room Fe2⫹F2 . Milling for 6 h 共Fig. 2b兲 results in further dehalogenation
关⬍1% relative humidity 共RH兲兴 for processing. Processing was per- and formation of the major phase of Fe2⫹F2 . Regarding the alterna-
formed in a dry room so as to impede any potential CMFNC reac- tive carbons for the CMFNCs, carbon black resulted in a small
tion with atmospheric moisture. Electrodes were prepared by adding degree of FeF2 at 4 h 共Fig. 2a兲 and an increasing amount at 6 h 共Fig.
poly共vinylidene fluoride-co-hexafluoropropylene兲 共Kynar 2801, Elf 2b兲. This is in contrast to the expanded graphite sample in which the
Atochem兲, carbon black 共Super P, MMM兲, and dibutyl phthalate FeF3 composition was maintained. The dehalogenation may be re-
共Aldrich兲 to the CMFNC in acetone. The slurry was tape cast, dried lated to adsorbed moisture, which is most prevalent in the activated
for 1 h at 22°C, and rinsed in 99.8% anhydrous ether 共Aldrich兲 to carbon, followed by the carbon black and then finally the expanded
extract the dibutyl phthalate plasticizer. The electrodes, 1 cm2 disks graphite. The contrast in efficiency of the dehalogenation reaction
typically containing 57 ⫾ 1% CMFNC and 12 ⫾ 1% carbon black among the carbons makes it unlikely that the formation of CFx is the
were tested electrochemically vs. Li metal 共Johnson Matthey兲. The product or is responsible for the resultant electrochemical properties.
Swagelock 共in-house兲 or coin 共NRC兲 cells were assembled in a He- This point is revisited in greater detail in the Discussion section.
filled dry box using Whatman GF/D glass fiber separators saturated Crystallite size was determined through application of the
with 1 M LiPF6 in ethylene carbonate:dimethyl carbonate 共EC:DMC Scherrer formula on the 共012兲 Bragg peak for CMFCS fabricated by
1:1 in vol兲 electrolyte 共Merck兲. The cells were controlled by Mac- milling FeF3 with the three carbons for 10, 30, 60, 240, and 360
A1320 Journal of The Electrochemical Society, 150 共10兲 A1318-A1327 共2003兲

Figure 2. XRD patterns of FeF3 :C CMFNCs formed from high-energy mill-


ing for 共a兲 4 and 共b兲 6 h using 共i兲 activated carbon, 共ii兲 carbon black, and 共iii兲
expanded graphite as the carbon. Milling times are indicated within each
figure. FeF2 phase is denoted by *, Fe2 F5 is denoted by ∧ .

min. The resulting data of the Scherrer calculation is plotted in Fig.


3. The 6 h activated-carbon-based CMFNC was not included due to
the small quantity of FeF3 present. All samples demonstrated similar
evolution of crystallite size with milling time. 10 min milling times
result in a large reduction of crystallite size from ⬎150 nm to ap-
proximately 40-45 nm. Further milling results in a minimum in crys-
tallite size of 25-30 nm between 60 and 240 min. Bragg peak broad-
ening, which is the basis of the Scherrer calculation, can also be
affected by nonhomogeneous internal strains which may not be
trivial in a high-energy-milled sample. The Williamson-Hall 共W-H兲
calculation is typically utilized to deconvolute the two contributing

Figure 4. DF TEM image at 共a兲 low and 共b兲 high magnification of 85/15 wt
% FeF3 /activated carbon CMFNC fabricated by high-energy milling for 1 h.
Bright areas associated with presence of FeF3 共see text兲.

factors. Due to poor peak intensity we were not able to run the W-H
calculation; therefore, TEM analysis was utilized to corroborate the
results of the Scherrer analysis.
The formation of a nanocomposite is supported by the DF TEM
image of a CMFNC fabricated by milling with activated carbon for
1 h 共Fig. 4a and b兲. The majority of the FeF3 domains 共bright areas
associated with diffracted intensity of the 共012兲 reflection of R3C
FeF3 ) were ⬍30 nm interdispersed with carbon. Further discussion
of the nanocomposite structure was previously presented9 and will
Figure 3. Crystallite size as a function of milling time for FeF3 :C CMFNCs be presented in a following publication. Unexpectedly, it was found
based on 共䊏兲 activated carbon, 共⽧兲 carbon black, and 共䊊兲 expanded graph- that further milling results in an increase in crystallite size, as is
ite. clearly seen in the trend established in Fig. 3.
Journal of The Electrochemical Society, 150 共10兲 A1318-A1327 共2003兲 A1321

Figure 5. Comparison of discharge/charge voltage profiles of FeF3 :C


CMFNCs 共1 h activated carbon兲 at 22 and 70°C at 7.58 mA/g. Data shows
first full discharge after short initial charge to 4.5 V.

The particle size increase could be attributed to a mechanically


assisted coarsening effect assisted by heat generated within the cell.
Another plausible explanation for this seemingly contradictory be-
havior 共increase in crystallite size with extended milling兲 can be
extracted from the examination of the XRD patterns. We have
shown CMFNCs composed of activated carbon have a greater trend
in crystallite size increase after 60 min of milling 共Fig. 3兲, followed
by the carbon black and the expanded graphite sample. The identical
trend 共activated carbon ⬎ carbon black ⬎ expanded graphite兲 is
also found in the carbon’s tendency to dehalogenate to the reduced
phases of Fe2 F5 and FeF2 共Fig. 2a and b兲. Through analysis of the
convoluted Bragg peak full width at half maximum 共fwhm兲, the
Scherrer formula presents a composite value for the primary particle
sizes and does not give an indication of distribution. From TEM
analysis, we have found that CMFNCs have a wide range of crys-
tallite size 共1-30 nm兲. One can envision the dehalogenation reaction
occurring most immediately for the smallest FeF3 crystallite sizes
共⬍5 nm兲, thereby leaving the remainder as larger FeF3 crystallites.
As a result of the lack of small crystallites, the Bragg peaks narrow
in the CMFNCs showing significant second-phase development 共de-
halogenation兲 and result in a false impression of growth in crystallite
size.
Electrochemical characterization 1.5-4.5 V.—The electrochemical
activity of the 85/15 wt % FeF3 :C CMFNCs was investigated in the
region between 4.5 and 1.5 V (Li/Li⫹) to explore the feasibility of a
reduction reaction from Fe3⫹ to Fe 共Fig. 5兲. Besides the initial Fe3⫹
to Fe2⫹ reaction in the voltage range between 4.5 and 2.5 V we
previously investigated,12 there exists a plateau development at ap-
proximately 2 V resulting in a total specific capacity of 367 mAh/g
共Fig. 5兲. This reaction was found to be reversible. The 2 V ‘‘plateau’’
seemed to exhibit Li insertion limitations as evidenced by the cur-
vature developing toward the end of reaction. Another cell was fab-
ricated and the electrochemical activity was investigated at 70°C to
increase kinetics of the reaction. At 70°C 共Fig. 5兲, the specific ca-
pacity of the CMFNC increased significantly to 660 mAh/g. Evi-
dence clearly showed that the process was at least partially revers- Figure 6. Voltage vs. time plots for Li/LiPF6 EC:DMC/FeF3 :C CMFNC 共1
ible with at least two distinct oxidation reactions occurring at h activated carbon兲 cells cycled at 70°C and 7.58 mA/g as a function of
approximately 3.0 V. In addition, the voltage vs. capacity plot also milling time.
showed that the 3.8-2.8 V (Fe3⫹ → Fe2⫹) reduction reaction that
was observed for the virgin FeF3 :C nanocomposite was reversible.
This, combined with the coulombic efficiency of the charge process shown in the representative voltage profiles of Fig. 6. The manually
with respect to the discharge, suggests complete reoxidation to milled sample exhibited very poor performance. A dramatic increase
Fe3⫹. in the electrochemical activity develops after only 10 min of milling
All the FeF3 -based CMFNC samples were initially discharged to to form the CMFNC and reaches an optimum for the 1 and 4 h
1.5 V and subsequently charged to 4.5 V at 70°C at 7.58 mA/g. The sample. After the third cycle, the discharge capacity was analyzed
effect of the milling time to form the CMFNC on the electrochemi- for all CMFNC samples. The data was separated into two regions,
cal behavior of the activated carbon-based CMFNC samples is the specific capacity recovered between 4.5 and 2.5 V, and the
A1322 Journal of The Electrochemical Society, 150 共10兲 A1318-A1327 共2003兲

Figure 9. Composite specific discharge and charge capacities for FeF3 :C


CMFNCs formed by milling for 6 h. Cells were cycled between 1.5 and 4.5
V at 7.58 mAh/g at 70°C.

Figure 7. Distribution of the third discharge specific capacity of FeF3 :C


CMFNCs as a function of carbon type and milling time. Cells were cycled at
7.58 mA/g at 70°C. average voltage is low 共approx. 2.3 V兲, energy densities based on the
smaller 540 mAh/g specific capacity number result in 1242 Wh/kg,
a factor of two greater than LiCoO2 ’s 600 Wh/kg. This is the high-
specific capacity recovered from 2.5 to 1.5 V. This would graphi- est specific capacity nonsulfur positive electrode reported to date. In
cally isolate the Fe3⫹ to Fe2⫹ and the Fe2⫹ to Fe0 reduction reac- addition, cycling efficiencies 共see charge/discharge specific capaci-
tions, respectively. These specific capacities 共based on weight of ties in Fig. 8 and 9兲 were very high, even at 70°C with high surface
complete nanocomposite兲 are shown in Fig. 7. Total CMFNC ca- area, highly active nanocomposites. Although specific capacities
pacities after the third cycle exceeded 540 mAh/g 共631 mAh/g were exceptionally high and efficiency was good, a noticeable fade
FeF3 ) in a number of samples. Approximately one third of the ca- in capacity with cycle number did exist. Preliminary evidence sug-
pacity was recovered in the upper voltage area between 2.5 and 4.5 gests that this condition can be addressed with improved electrolytes
V. This behavior is consistent with the 1e⫺ process at high potential and attention to electrolyte/active material interface.
vs. the 2e⫺ process at low potential. Comparison of the development
of the specific capacity of the samples 共Fig. 7 and 6兲 with the crys- Discussion
tallite size 共Fig. 3兲 clearly shows a strong and direct relationship.
The development of the extraordinary specific capacity is tied di- Discharge reaction mechanism.—Electrochemistry and
rectly to the formation of a nanocomposite, as we have previously mechanochemistry.—With capacities reaching 712 mAh/g, the the-
shown in detail for the 3 V reaction in FeF3 -based CMFNCs.9 oretical reduction capacity of FeF3 , there is little doubt that full
Specific capacity as a function of cycle number for the CMFNCs reduction of Fe3⫹ to Fe0 is occurring. The Fe3⫹ to Fe2⫹ reaction
milled for 4 and 6 h is shown in Fig. 8 and 9, respectively. The mechanism was found to consist of a two-phase transformation fol-
highest capacities recovered were for the 6 h carbon black-based lowed by a topotactic insertion reaction as suggested by in situ and
sample with a second discharge capacity of 660 mAh/g for the com- ex situ XRD and electrochemical characterization 共Eq. 2兲. Therefore,
plete composite 共15% carbon black兲. Corrected for the weight of as we enter the reaction of discussion at 2 V, we begin with LiFeF3 .
FeF3 one gets 759 mAh/g. This agrees well with the theoretical Reducing the Fe2⫹ in LiFe2⫹Fe3 to Fe0 would necessitate the break-
capacity of FeF3 共712 mAh/g兲. The slight excess in capacity could ing of all Fe-F bonds and result in significant structural rearrange-
be attributed to irreversible parasitic cathodic reactions occurring on ment. We analyzed by ex situ XRD the reaction products of the fully
the carbon surface. This effect dissipated within two full cycles. lithiated sample to elucidate the reaction mechanism.
Most samples cycled in the 540 mAh/g range resulting in a cor- A FeF3 :C 1 h activated carbon CMFNC was reduced to 1.5 V.
rected specific capacity of 621 mAh/(g FeF3 ). These capacities are The powder was removed in a He-filled glove box and washed in
approximately 400% greater than present day LiCoO2 . Even though DMC to remove excess salts. The powder was then analyzed by
XRD in a sealed glass slide to minimize atmospheric exposure. The
results are shown in Fig. 10. Two compounds were identified, LiF

Figure 8. Composite specific discharge and charge capacities for FeF3 :C Figure 10. XRD patterns of ex situ analysis of reaction products postelec-
CMFNCs formed by milling for 4 h. Cells were cycled between 1.5 and 4.5 trochemical and mechanochemical reduction of FeF3 :C CMFNCs with Li⫹
V at 7.58 mAh/g at 70°C. or Li 共see text兲.
Journal of The Electrochemical Society, 150 共10兲 A1318-A1327 共2003兲 A1323

Figure 12. Ex situ XRD of FeF3 :C CMFNCs 共activated C 4 h兲 at various


states of charge. 共*兲 Refers to developing peaks.

FeF3 -based CMFNC, namely, a 1e⫺ reaction between 3.5 and 2.8 V
and a 2e⫺ reaction at approximately 2 V. Based on this limited
information we can state that it is likely the Fe0 is reoxidizing to
Fe3⫹. However, we cannot make a statement regarding the reconsti-
tution of the FeF3 phase in a reversible fluoride conversion process.
Ex situ XRD characterization.—During the conversion reaction at 2
V the CMFNC converted into an even finer nanocomposite of Fe
Figure 11. In situ XRD of FeF3:C CMFNC 共1 h activated carbon兲. Note and LiF on the scale of 10-20 Å based on XRD and HRTEM analy-
systematic decrease in 共104兲, 共110兲, and 共113兲 Li1 FeF3 intensity and the sis 共discussed later兲. During oxidation, in situ XRD showed little
corresponding increase in 共110兲 peak of Fe. 共.....兲 Placed as an aid to the eye. useful information except for the systematic decrease in the intensity
of the broad Bragg peak associated with disappearance of Fe. To
gain more insight into the reaction, the 4 h milled FeF3 :C CMFNC
and Fe. This is consistent with a thermodynamic conversion reaction utilizing activated carbon electrodes were made into three cells in
of the type which one was discharged to 1.5 V, one recharged to 3.3 V, and one
recharged to 4.5 V at 70°C. The three cells were then disassembled
2Li⫹ ⫹ 2e⫺ ⫹ LiFe2⫹F3 → 3LiF ⫹ Fe 关3兴 in a He-filled glove box and analyzed by XRD. The results are
shown in Fig. 12. The fully lithiated sample consisted of Fe and LiF
The thermodynamics of such a reaction is consistent with the volt- as in the previous sample described previously. During oxidation,
age at which the reaction was found to occur, as discussed shortly. the foremost feature is the reduction in the intensity of the Bragg
To further support the mechanism, the electrochemical reaction peak at approximately 2.0 and 2.3 Å, associated with the disappear-
was replicated through the use of mechanochemistry. Li metal was ance of the Fe and LiF phase, respectively. Also apparent is the
placed in a mechanomilling cell with an activated carbon:FeF3 CM- development of weak peaks at approximately 1.7, 1.8, 2.7, and 3.3
FNC in a 3:1 molar ratio under a He atmosphere. The mechanical Å. Few of these peaks can be associated with the original FeF3
mill was run for 30 min and the sample was removed in the glove phase or the Li1 FeF3 phase present before conversion. Although of
box and put on a sealed glass slide for XRD analysis. The mecha- weak intensity, a striking agreement of the peaks was found with
nochemical results are plotted vs. the electrochemical results 共Fig.
FeF2 . This is shown by comparing the delithiated sample with a
10兲. The results are virtually identical, the formation of Fe and LiF,
therefore supporting the theory of a thermodynamically driven con- FeF2 CMFNC formed by milling FeF2 with activated carbon for 1 h
version mechanism. This conversion, although destructive in nature, in a 85:15 wt % stoichiometry 共Fig. 13兲. It should be noted that most
coincides with a relatively small 16% volume expansion and there- of the ‘‘FeF2 ’’-like phase formation occurs by the 3.3 V scan.
fore should not preclude its use in an electrochemical cell. Ex situ SAED characterization.—The ex situ XRD data used in the
In situ XRD.—The development of the conversion reaction was previous analysis was of poor intensity. In addition, the 4 h sample
characterized by in situ XRD in order to determine whether an in- initially contained nontrivial quantities of FeF2 , which may have
termediate reaction takes place during the conversion reaction. The reconverted. To confirm the phase development upon oxidation for
initial lithiation reaction 共Reaction 2兲 to LiFeF2 occurring between
3.5 and 2.8 V has been previously characterized in the in situ analy-
sis down to 2.5 V.9 Figure 11 shows the in situ analysis from 2.5 to
1.5 V at 22°C. From x ⫽ 0.92 to 2.42 in Lix FeF3 , all the peaks
associated with the Li1 FeF3 phase systematically decrease in inten-
sity 关especially for the 共113兲 peak兴. Simultaneously, the development
of the Fe 共110兲 peak is found to systematically increase. Although
intensities are poor, these results suggest the final 2e⫺ transfer pro-
cess from Fe2⫹ to Fe0 proceeds in a straightforward two-phase con-
version reaction as described in Eq. 3.
Oxidation reaction and reversibility.—We have come to a deter-
mination of the mechanisms involving the reduction reaction. The
question remains as to what is the reaction mechanism of the oxi-
dation reaction supporting the reversibility. As discussed before, the
oxidation reaction is associated with approximately 3e⫺ transfer and Figure 13. Ex situ XRD of FeF3 :C CMFNCs 共activated C 4 h兲 at 共a兲 full
subsequent discharge reactions maintain characteristics of the initial charge compared with 共b兲 raw FeF2 :C CMFNC 共activated C 1 h兲.
A1324 Journal of The Electrochemical Society, 150 共10兲 A1318-A1327 共2003兲

Figure 14. SAED images taken from


two fields each of FeF3 :C CMFNCs that
were discharged to 1.5 V (Li2.58FeF3 )
and recharged to 4.5 V (Li0.596FeF3 ) all
at 70°C. Major diffraction circles are la-
beled with their respective d-spacings in
共Å兲. Data from Tables I and II were de-
rived from these spectra.

the FeF3 CMFNC with more certainty, SAED, which has a higher fully discharged sample 共Table I兲. Analysis shows they agree well
sensitivity than X-rays, was utilized to characterize electrochemi- with FeF2 and the previously described ex situ XRD data.
cally formed samples. The 1 h milled FeF3 :C CMFNC utilizing Lattice parameters were calculated on the basis of the
activated carbon was made into four cells. One was discharged to P42/mnm FeF2 structure. The peak at approximately 2.0共3兲 Å was
2.2 V (Li1.23FeF3 ), one discharged to 1.5 V (Li2.58FeF3 ), one re- associated to a composite of residual Fe from lack of full reconver-
charged to 3.1 V (Li2.18FeF3 ), and one recharged to 4.5 V sion, and LiF which would be left as a 1 mol constituent 共see Eq. 4兲,
(Li0.596FeF3 ), all at 70°C. Upon analysis, the SAED images and and finally the 共210兲 Bragg reflection from FeF2
resulting d-spacings of the samples portray striking contrasts, espe-
cially between the ones taken of the 1.5 and 4.5 V samples 共Fig. 14兲. 3LiF ⫹ Fe → FeF2 ⫹ LiF 关4兴
The sample at 2.2 V 共not shown兲 was observed to be mostly nano-
crystalline with crystals a few nanometers in size, as evidenced by The 共220兲 data point was not included in the calculation due to
broad and diffuse diffraction rings. This was because ⬎1e⫺ was potential convolution with residual LiF and/or Fe diffraction rings
transferred and the conversion reaction to LiF and Fe was underway. 共as explained later兲. The lattice parameter data 共Table I兲 showed a
The fully lithiated 1.5 V sample showed diffuse rings associated to good fit to an FeF2 -like structure with a slightly smaller c/a ratio
the full conversion product Fe and LiF 共Fig. 14兲; the nanocrystalline with respect to the JCPDS standard.
nature of the composite is confirmed in the HRTEM image of the The only FeF2 peak which did not develop to a great extent in
discharged composite shown in Fig. 15a. d-spacings derived from the reconverted samples is the 共110兲 associated with the ordering of
SAED spectra taken from two fields of the 1.5 V sample are shown Fe in the 关110兴. This may be due to orientation of the domains
in Table I. Field 1 had d-spacings which all agreed well with that of within the SAED spot area or interlayer mixing of the Fe cations
LiF and possibly small quantities of Fe. The second field had upon reconversion. The latter could play a role in the decrease of the
d-spacings which agreed well with that of the Fe majority. Both c/a ratio. The ex situ XRD analysis of Fig. 13 showed that although
results were consistent with that of the ex situ powder XRD results attenuated, the peak was present. We are investigating in more detail
and the conversion reaction of Eq. 3. the root of the attenuated 共110兲 reflection and the crystal chemistry
The partial recharge to 3.1 V 共not shown兲 and full recharge to 4.5 of the reconverted phase.
V showed new rings associated with new d-spacings with a parallel We have shown that XRD and, even more so, the SAED evi-
development in crystallinity as represented by more spot develop- dence supports the reformation of an iron fluoride phase from the
ment on the SAED pattern 共Fig. 14兲. The HRTEM image of Fig. LiF and Fe cathodic sweep reaction products 共Eq. 3兲. However,
15b, although of limited field of view, also supports this trend. List- evidence supports that the reconverted phase is related in structure
ing of the d-spacings of the SAED spectra from the sample re- to Fe2⫹F2 , not Fe3⫹F3 as we expected. This revelation is com-
charged to 4.5 V is shown in Table II for two different fields of pounded with the added discrepancy that subsequent cycles show
analysis. The peaks from both fields differ significantly from the that close to 3e⫺ are reversibly consumed in the course of the
Journal of The Electrochemical Society, 150 共10兲 A1318-A1327 共2003兲 A1325

Table II. d-spacings derived from SAED analysis of two fields of


a FeF3 C CMFNC recharged to 4.5 V. JCPDS standard for FeF2 is
shown for reference. Lattice parameters are indexed to FeF2
structure; „210… was not utilized in calculation due to potential
convolution with Fe and LiF diffraction rings.

4.5 V Field 1 4.5 Field 2 FeF2 - P42/mnm


3.320 共110兲
2.661 2.694 2.700 共101兲
2.339 2.377 2.344 共111兲
2.034 2.014 2.101 共210兲
LiF 2.013 共220兲
Fe 2.014 共110兲
1.792 1.773 共211兲
1.706 1.660 共220兲
1.530 1.541 1.488 共310兲
1.482 共112兲
1.433 1.433 1.417 共310兲
1.219 1.213 共321兲
1.165 1.160 1.173 共222兲
a ⫽ 4.79 ⫾ 0.02 Å a ⫽ 4.85 ⫾ 0.03 Å a ⫽ 4.7 Å
c ⫽ 3.20 ⫾ 0.05 Å c ⫽ 3.15 ⫾ 0.03 Å c ⫽ 3.31 Å

form a FeF2 -based CMFNC. The material’s electrochemistry was


then evaluated at 70°C. This material also developed a reversible
3.5-2.8 V feature with comparable specific capacity 共620 mAh/g
based on metal fluoride weight only, Fig. 16兲 as compared to the
FeF3 -based CMFNC. This feature only develops after the conver-
sion process is completed; as such, the second discharge profile is
shown in comparison with the FeF3 CMFNC

FeF2 ⫹ 2Li ⫹ 2e⫺ → 2LiF ⫹ Fe 关5兴

The electrochemical feature at 3 V validates the possibility that FeF2


could be responsible for such a feature, but first one must be certain
that it is iron fluoride, which is the root of such a reaction.
The presence of a metal fluoride and carbon under high-energy
Figure 15. Ex situ high-resolution TEM images of FeF3 :C CMFNCs that
conditions may lead one to speculate that CFx can be formed. Al-
were 共a兲 discharged to 1.5 V (Li2.58FeF3 ) and 共b兲 recharged to 4.5 V
though traditionally not a reversible reaction, the 3 V range of inter-
(Li0.596FeF3 ) all at 70°C.
est in the reactions is consistent with the cathodic reaction identified
for a range of CFx compounds synthesized at low temperature. How-
ever, the evidence clearly suggests that CFx compounds play no role
electrochemistry as opposed to 2e⫺ that we expect for an FeF2 in the electrochemical activity explored in this manuscript.
phase reconversion. In addition, the reversible profile clearly shows The first evidence is found in the investigation of the metal fluo-
the 3.5-2.8 V region associated with the Fe3⫹ → Fe2⫹ reduction ride dehalogenation reaction which can easily be associated with the
along with the conversion at approximately 2 V associated with the formation of carbon fluorides
Fe2⫹ → Fe0 reduction. The former reduction reaction would not
be expected to occur in Fe2⫹F2 . In the following paragraphs we MeFx ⫹ C → CFx ⫹ Me 关6兴
enter an introductory discussion to try to explain this phenomena
and to rule out other hypothetical situations which can give rise to
such behavior.
As a first step, we investigated whether nanocomposites utilizing
FeF2 would show such behavior. Therefore, 85 wt % FeF2
⫺ 15 wt % activated carbon were high-energy milled for 1 h to

Table I. d-spacings derived from SAED analysis of two fields of a


FeF3 :C CMFNC discharged to 1.5 V. JCPDS standards for LiF
and Fe shown for reference.

1.5 V Field 1 LiF-Fm3m 1.5 V Field 2 Fe-Im3m


2.301 2.325 共111兲
2.011 2.013 共200兲 2.014 2.027 共110兲
1.420 1.424 共220兲 1.429 1.433 共200兲
1.207 1.214 共311兲
1.156 1.163 共222兲 1.163 1.170 共211兲 Figure 16. Comparison of second cycle voltage curves for FeF3 :C and
1.003 1.005 共400兲 1.001 1.013 共220兲 FeF2 :C CMFNCs showing distinct similarities. All cells were cycled at 7.58
0.922 共331兲 0.898 0.906 共310兲 mA/g at 70°C in LiPF6 EC:DMC and were precharged to 4.5 V.
A1326 Journal of The Electrochemical Society, 150 共10兲 A1318-A1327 共2003兲

Figure 18. FeF3 -based nanocomposites utilizing 15 and 25 wt % V2 O5 as


conductive matrix. Cell cycled at 7.58 mA/g at 70°C. Specific capacity is
based on weight of entire nanocomposite.

redox of the Fe transition metal. Within an FeF2 structure we have


behavior which is consistent with an Fe2⫹ → Fe3⫹ transition. One
possible mechanism would be the formation of FeOx F2⫺2x by elec-
trochemical decomposition of solvents. However, the covalency
brought about by the oxygen should evolve a separate redox reaction
which is approximately 1 V less than the corresponding fluoride
conversion reaction 共as shown in following section兲, thereby ruling
out this hypothetical mechanism.
Figure 17. Voltage vs. time for 共a兲 NiF2 :C and 共b兲 CoF2 :C 85:15 wt % One cannot ignore the effect of surface area and the associated
CMFNCs cycled at 7.58 mA/g at 70°C. ‘‘Spikes’’ in voltage were due to surface reactions. The reconversion of the cathodically formed
unintentional periods of open-circuit condition which portray the poor kinet- 3LiF ⫹ Fe to anodically formed FeF2 leads to a nanocomposite
ics of the electrochemical conversion reaction. consisting of nanograins on the order of 5 nm. It is impossible to
estimate the true metal fluoride/electrolyte area of the reconverted
material, although it is not unrealistic to speculate a large percentage
It was observed that significant dehalogenation occurs with extended of the grains have active interfaces. This high-surface-area compos-
milling times. Therefore, if the dehalogenation reaction is associated ite could enable the oxidation of Fe2⫹F2 to Fe3⫹F2 by a pseudoca-
with the formation of electrochemically active CFx , we would ex- pacitive reaction. This reaction would consist of the PF⫺ 6 anion of
pect specific capacity to increase drastically as the milling times are the salt adsorbed to form a surface entity Fe3⫹Fe2 (PF⫺ 6 upon the
)
increased. This is not the case as Fig. 7 clearly shows the specific anodic sweep. The pseudocapacitive behavior as an active mecha-
capacity of the partly dehalogenated 共Fig. 2兲 6 h milled activated nism for the FeF2 oxidation is too lengthy to discuss in depth here
carbon sample decreases. Within the same argument, the graphite- and will be discussed in more detail in the next paper of this series.
based samples which exhibit very little dehalogenation 共Fig. 2兲 re- In short, it is a viable explanation for 共i兲 the additional specific
veal very high specific capacity 共Fig. 7兲. capacity and 共ii兲 the 3.5-2.8 V voltage feature developing in the
The second argument is that NiF2 and CoF2 have more oxidizing FeF2 phase upon reconversion of the FeF3 :C CMFNC.
power and should promote CFx formation and the 3 V profile if the
two were related. To prove this point we fabricated NiF2 and CoF2 Thermodynamics and ionicity.—The conversion reaction of the
CMFNCs utilizing an activated carbon matrix. Their electrochemi- Li1 FeF3 → 3LiF ⫹ Fe (Fe2⫹ → Fe0 ) reaction is consistent with a
cal activity is shown in Fig. 17, which is in sharp contrast to the thermodynamically driven reaction. Assuming a Gibbs free energy
activity of FeF2 CMFNC 共Fig. 16兲, namely, in the absence of the of Fe2⫹F2 to be approximately equivalent with that of LiFe2⫹F3 the
3.5-2.8 V reduction reaction. calculation of the free energy of Eq. 3 is 2.44 V. Experimentally, a
The evidence and subsequent arguments presented suggest that slightly lower potential 共approximately 2 V兲 was found for the re-
CFx is not the origin of the electrochemical features observed in the action to proceed. This may be nestled in the slow kinetics of the
reaction as suggested by its temperature dependence. In addition, the
FeF3 -based CMFNCs. To be more conclusive in our determination,
we have removed carbon entirely from the nanocomposite system lithiated iron trifluoride is different in lattice energy than the FeF2 on
which the thermodynamic values of the calculation were based. The
and replaced it with an electrically conductive oxide. V2 O5 was
thermodynamics of Reaction 7 is interesting
high-energy milled in 15 and 25 wt % quantities with FeF3 . Physi-
cal and electrochemical characteristics will be discussed in detail in Li⫹ ⫹ e⫺ ⫹ FeF3 → LiF ⫹ Fe 关7兴
a future publication; however, Fig. 18 clearly shows that the vana-
dium oxide-based nanocomposite enabled the electrochemical activ-
If the conversion reaction could occur directly from Fe3⫹ to Fe0 , the
ity of FeF3 in the same fashion as carbon 共Fig. 16兲. Therefore, in no
theoretical output voltage would be 2.74 V at a capacity of 712
way can CFx be associated with the electrochemical characteristics mAh/g 共1950 Wh/kg兲, thereby eliminating the unattractive two-step
of the fluoride nanocomposites.
process we have witnessed. However, even if the Fe3⫹ to Fe2⫹
Now that CFx has been removed as a candidate for the observed lithium insertion process could be overcome, we would still be faced
electrochemical activity, we now attempt to explain the 3 V feature with a two-step thermodynamically driven conversion process
which is found both in the FeF2 -like material formed postconversion
of FeF3 , and from FeF2 itself. Li⫹ ⫹ e⫺ ⫹ FeF3 → LiF ⫹ FeF2 2.90 V 关8兴
The key to the explanation of the additional specific capacity and
the 3 V feature is to go to the root of the question which lies in the 2Li⫹ ⫹ e⫺ ⫹ FeF2 → 2LiF ⫹ Fe 2.44 V 关9兴
Journal of The Electrochemical Society, 150 共10兲 A1318-A1327 共2003兲 A1327

Li⫹ ⫹ e⫺ ⫹ FeF3 → LiFeF3 关14兴

2Li ⫹ 2e⫺ ⫹ LiFeF3 → 3LiF ⫹ Fe 关15兴

In summary, it is surprising that the fluoride-based reaction is


reversible. The potentials utilized during the anodic reaction are
much higher than the analogous oxide reaction. They exceed the
values to oxidize the Fe into the solution of the electrolyte.

Conclusions
The reversible conversion reaction of metal fluorides was en-
abled through the use of nanocomposites 共CMFNCs兲. The ability to
reversibly reduce to the metal and reform a metal fluoride opens a
new avenue to high-specific-capacity positive electrodes based on
the reversible conversion reaction for lithium battery applications.
Clearly this can be extended to other classes of highly ionic,
transition-metal-based materials.
At this early stage, the high capacities 共⬎600 mAh/g兲 of the
fluorides are only accessible at elevated temperature and/or low rate.
These issues will no doubt be addressed by attention to kinetics and
Figure 19. Discharge curve of FeF3 :C CMFNC 共activated C 1 h兲 compared materials engineering, leading to more practical methods of nano-
with that of Fe2 O3 . Both cells were discharged at 7.58 mA/g at 70°C. composite production. However, based on the experimental data, we
believe the combination of fluoride and nanocomposite technology
offers an exciting avenue into the search for lithium rechargeable
Recently, the reversible conversion process of transition metal chal- batteries with energy densities double of today’s technology.
cogenides has been investigated by Poizot et al.8 and by our group
for nitrides.9 Due to the covalency of the chalcogenides and nitrides Acknowledgments
the conversion reaction occurs at 1 V potential. The ionicity of the G.G.A. sincerely thanks the U.S. Government for its support of
Fe-F bond, compared with the Fe-O bond, which is of more covalent this research. F.C. thanks P. Stodilman of EPFL Switzerland for use
character, results in a higher reduction potential, approximately of the electron microscopy facility.
equivalent to that of Fe2⫹ ions in solution. This is exemplified in
Telcordia Technologies assisted in meeting the publication costs of this
Fig. 19 which compares the second discharge of the electrochemical
article.
conversion of Fe3⫹ 3⫹
2 O3 with that of Fe F3 . Overall, there is an
approximate 1 V difference between the two conversion reactions. References
This is excellent agreement with the potential as calculated utilizing 1. K. Mizushima, P. C. Jones, P. J. Wiseman, and J. B. Goodenough, Mater. Res. Bull.,
the Gibbs free energy of reaction for the following discharge reac- 15, 783 共1980兲.
tion 2. M. G. S. R. Thomas, W. I. F. David, J. B. Goodenough, and P. Groves, Mater. Res.
Bull., 20, 1137 共1985兲.
6Li⫹ ⫹ 6e⫺ ⫹ Fe2 O3 → 3Li2 O ⫹ Fe 共 1.63 V兲 关10兴 3. M. M. Thackeray, W. I. F. David, P. G. Bruce, and J. B. Goodenough, Mater. Res.
Bull., 18, 461 共1983兲.
⫹ ⫺
3Li ⫹ 3e ⫹ FeF3 → 3LiF ⫹ Fe 共 2.74 V兲 关11兴 4. J. M. Tarascon, E. Wang, F. K. Shokoohi, W. R. McKinnon, and S. Colson, J.
Electrochem. Soc., 138, 2859 共1991兲.
5. A. R. Armstrong and P. G. Bruce, Nature (London), 381, 499 共1996兲.
Examination of Fig. 19 reveals intrinsic similarities between the 6. A. K. Padhi, K. S. Nanjundaswamy, and J. B. Goodenough, J. Electrochem. Soc.,
voltage curves associated with these reactions sans the 1 V differ- 144, 1188 共1997兲.
ence. During the Fe3⫹ to Fe2⫹ reduction reaction in the second 7. A. K. Padhi, K. S. Nanjundaswamy, C. Masquelier, S. Okada, and J. B. Good-
enough, J. Electrochem. Soc., 144, 1609 共1997兲.
cycle, both reactions proceed in a single phase-like reaction span- 8. P. Poizot, S. Laruelle, S. Grugeon, L. Dupont, and J. M. Tarascon, Nature (Lon-
ning a 1 V voltage range. The similarity extends to the Fe2⫹ don), 407, 496 共2000兲.
→ Fe0 reaction. For both the chalcogenide and halide, this reaction 9. N. Pereira, L. C. Klein, and G. G. Amatucci, J. Electrochem. Soc., 149, A262
proceeds in a two-phase conversion reaction reducing the divalent 共2002兲.
10. M. Kamiya, M. Eguchi, T. Miura, and T. Kishi, Solid State Ionics, 109, 321 共1998兲.
iron to metallic iron. The similarity of the reaction is summarized by 11. H. Arai, S. Okada, Y. Sakurai, and J. Yamaki, J. Power Sources, 68, 716 共1997兲.
the following equations 12. F. Badway, N. Pereira, F. Cosandey, and G. G. Amatucci, J. Electrochem. Soc., 150,
A1209 共2002兲.
2Li ⫹ 2e⫺ ⫹ Fe2 O3 → Li2 Fe2 O3 关12兴 13. J. Maier, Solid State Ionics, 148, 367 共2002兲.
14. J. L. Lábár, in Proceedings of EUREM 12, July 2000, Brno, L. Frenk and F.
4Li ⫹ 4e⫺ ⫹ Li2 Fe2 O3 → 2Li2 O ⫹ Fe 关13兴 Ciompor, Eds., Vol. III, p. 1379-1380 共2000兲.

View publication stats

You might also like